Return-Path: <john.harrison-request@uk.ac.cam.cl>
Received: from ted.cs.uidaho.edu by swan.cl.cam.ac.uk with SMTP (PP-6.0)
          id <20528-0@swan.cl.cam.ac.uk>; Fri, 13 Mar 1992 09:42:52 +0000
Received: by ted.cs.uidaho.edu (16.6/1.34) id AA13156;
          Fri, 13 Mar 92 01:31:44 -0800
Sender: info-hol-request@edu.uidaho.cs.ted
Errors-To: info-hol-request@edu.uidaho.cs.ted
Received: from swan.cl.cam.ac.uk by ted.cs.uidaho.edu (16.6/1.34) id AA13152;
          Fri, 13 Mar 92 01:31:38 -0800
Received: from aldham.cl.cam.ac.uk by swan.cl.cam.ac.uk
          with SMTP (PP-6.0) to cl id <20111-0@swan.cl.cam.ac.uk>;
          Fri, 13 Mar 1992 09:34:17 +0000
Received: by aldham (5.57/SMI-3.0DEV3) id AA05143; Fri, 13 Mar 92 09:34:13 GMT
Date: Fri, 13 Mar 92 09:34:13 GMT
From: John.Carroll@uk.ac.cam.cl
Message-Id: <9203130934.AA05143@aldham>
To: info-hol@edu.uidaho.cs.ted
Cc: John.Carroll@uk.ac.cam.cl
Subject: HOL2.0 port to PC-compatibles


I have recently finished porting HOL2.0 to PC-compatibles running
Windows and Procyon Common Lisp. Performance looks to be good, for
example a 33MHz 486 running faster than either a Sparc SLC or IPC
under Allegro CL or AKCL. Here are some comparisons of run time for
the multiplier benchmark:

      Machine        Lisp            Memory   Window System   Total Time
   -----------+--------------------+--------+---------------+-----------
   33MHz 486  | Procyon            |  8Mb   | Windows 3.0   | 151.9 sec
   PC clone   |                    |        |               |

   -----------+--------------------+--------+---------------+-----------
   Sparc SLC  | Allegro            |  16Mb  |    X (R4)     | 214.4 sec
   -----------+--------------------+--------+---------------+-----------
   Sparc SLC  | AKCL               |  16Mb  |    X (R4)     | 278.2 sec
   -----------+--------------------+--------+---------------+-----------
   Sparc IPC  | Allegro            |  16Mb  |    X (R4)     | 174.9 sec
   -----------+--------------------+--------+---------------+-----------
   Sparc IPC  | AKCL               |  16Mb  |    X (R4)     | 235.2 sec
   -----------+--------------------+--------+---------------+-----------
   SparcServ  | Lucid              |  64Mb  |    NONE       | 102.7 sec
   -----------+--------------------+--------+---------------+-----------
   SparcServ  | AKCL               |  64Mb  |    NONE       | 238.4 sec
   -----------+--------------------+--------+---------------+-----------
   Sparc2     | Allegro            |  32Mb  |   SunView     | 124.1 sec
   -----------+--------------------+--------+---------------+-----------
   Sparc2     | AKCL               |  32Mb  |   SunView     | 148.0 sec
   -----------+--------------------+--------+---------------+-----------
   DS5000/200 | KCL                |  48Mb  |    NONE       | 287.6 sec
   -----------+--------------------+--------+---------------+-----------
   Mac IIci   | Procyon            |  8Mb   |  Mac. Win.    | 359.6 sec
   -----------+--------------------+--------+---------------+-----------

I haven't had a chance to run the whole benchmark on a 386 machine, but
extrapolating from running just the first part of it in a beta-release
version of Procyon I estimate that an average 25MHz 386 would end up
around the 320 second mark.

Some details: the machine needs to be a 80386 or 80486 (may be SX or
DX) with at least 5MB or memory -- preferably 8MB -- running Microsoft
Windows 3 in 'enhanced mode'. Also needed is a copy of Procyon Common
Lisp (costing 450 pounds sterling educational / 1500 pounds industry,
obtainable from the address given below). The HOL system image occupies
around 5MB disc space, and another 2MB is required for a minimal set of
theory files etc.

Judging from the benchmark figures, PC-compatibles look to be a
cost-effective way of getting raw HOL power. (The 486 machine that I
used cost under 2000 pounds).  They also provide an inexpensive way
(together with the new Macintosh notebook computers) of using HOL on a
portable computer.  One caveat, though, is that these are not UNIX
machines so shell scripts, Makefiles etc will not work; in particular,
building the HOL system from scratch requires a fair amount of manual
intervention.

Please e-mail me if you are interested in either a copy of the patches to
HOL2.0 that are necessary and/or a ready-built copy of the HOL system
image to go with a copy of Procyon that you are buying.


John Carroll
Cambridge University Computer Lab


Procyon Common Lisp is sold by: Scientia Ltd., St. John's Innovation
Centre, Cowley Road, Cambridge, CB4 4WS, UK; phone +44-223-421221,
fax +44-223-421218.


